Creator Platform Evidence Schema

Overview

The creator-foundation pack (formerly creator-media) currently has strong YouTube evidence workflows and platform-neutral strategy skills. The next expansion should start with a platform capability matrix and shared evidence schema before adding more platform-specific audit skills.

The first practical use case is LinkedIn plus a repo-maintained creator presence dossier: a Markdown record of a creator's public professional footprint, career arc, proof assets, platform roles, and content opportunities. The collection model must use free, open-source, and manual evidence paths by default. Paid APIs, privileged platform programs, and logged-in scraping are out of scope for the baseline.

Goals

Define a platform capability matrix that tells future skills what can be collected, how reliable it is, and what audit depth it supports.
Define a shared creator evidence schema that can normalize posts, videos, newsletters, podcasts, talks, repos, product proof, and career milestones.
Establish free/manual collection methods: owner exports, RSS/feeds, public page capture, manual snapshots, open-source extraction tools, and user-provided files.
Make LinkedIn the first supported non-YouTube evidence lane using owner export and public/manual capture.
Specify a creator-presence-dossier artifact that can live in a git repo and evolve over time.
Preserve the current evidence-first standard: cite raw paths, source URLs, capture time, metric gaps, and confidence limits.

Non-Goals

Do not build paid API integrations as the default path.
Do not require platform developer approval before a creator can use the pack.
Do not scrape logged-in, paywalled, private, or access-controlled surfaces.
Do not bypass bot protections, rate limits, or platform access controls.
Do not promise YouTube-equivalent transcripts, thumbnails, or performance metrics for every platform.
Do not mutate external accounts, publish content, delete content, or change creator profiles.

Detailed Design

Skill Set

Add these skills before adding more platform-specific audits:

Skill	Purpose	Output
`creator-platform-capability-matrix`	Map each platform to available free/manual evidence sources, fields, risks, and audit depth.	`research/creator-platforms/capability-matrix.md`
`creator-evidence-schema`	Normalize evidence fields and define raw evidence folders for future audits.	`research/creator-platforms/evidence-schema.md`
`creator-presence-dossier`	Maintain a repo-based Markdown dossier of the creator's career, public presence, platform roles, proof assets, and strategy implications.	`research/creator-presence/<slug>.md`

The existing platform-neutral skills should continue to consume these artifacts:

creator-positioning
content-programming
series-spec
product-led-media-map
creator-metrics-review

Collection Methods

Each evidence item must declare one collection method:

Method	Description	Baseline status
`export`	Owner-provided platform export, CSV, ZIP, or analytics download.	Preferred for LinkedIn and private metrics
`manual_snapshot`	User saves a page, screenshot, CSV, Markdown note, or copied post text.	Preferred when platform access is restricted
`rss_feed`	RSS, Atom, JSON feed, podcast feed, sitemap, or newsletter feed.	Preferred where available
`public_page_capture`	Public unauthenticated page capture with polite retrieval.	Allowed for public pages
`open_source_tool`	OSS tooling such as `trafilatura`, Playwright screenshots, `yt-dlp`, or feed parsers.	Allowed when ToS/access constraints are respected
`free_api`	Official free API or public protocol with no paid dependency.	Allowed, not required

Open-Source Tooling

The baseline should not install dependencies automatically. Skills may recommend optional tools and stop if missing:

trafilatura for extracting readable article/page text and metadata from public or locally saved HTML.
Playwright for public screenshots and HTML capture when a visual record matters.
Feed parsers for RSS/Atom/JSON feeds.
yt-dlp only for YouTube or other supported public media where the existing project policy permits it.
gallery-dl only for media/archive use cases where the platform and creator rights make that appropriate.

Platform Capability Matrix

The matrix should classify each platform by:

Platform and content types.
Evidence sources: export, manual snapshot, RSS/feed, public page capture, OSS tool, free API.
Fields likely available.
Fields likely unavailable.
Whether metrics are public, owner-only, or unavailable.
Whether content body is collectible.
Whether media assets are collectible.
Whether transcripts are available.
Whether peer benchmarking is practical.
Operational risk: low, medium, high.
Recommended first skill, if any.

Initial platform rows:

Platform	Baseline lane	Audit depth
LinkedIn personal profile	Owner export plus manual/public snapshots	Presence, career arc, post history, topic patterns; limited performance metrics
LinkedIn company page	Manual/admin export first; free API later only if already authorized	Page content and organization-level analytics when available
Personal website/blog	RSS/sitemap/public capture	High-quality body text, metadata, content themes
Newsletter	Export/RSS/public archive	Posts, cadence, themes, subscriber metrics only if owner-provided
Podcast	RSS/public show pages/manual transcripts	Episodes, guests, titles, cadence, descriptions, transcripts only if available
GitHub	Public API/manual repo evidence	Projects, proof assets, contribution narrative
X/Threads/Instagram/TikTok	Manual snapshots/exports first	Presence and packaging; metrics only if owner-provided
Bluesky/Mastodon	Public protocol/API where available	Public posts, engagement counters, topic patterns

Shared Evidence Schema

Raw evidence should live under:

research/creator-platforms/data/<platform>/<slug>/

Normalized evidence records should use JSONL when machine-generated and Markdown tables when manually curated.

Core fields:

Field	Meaning
`evidence_id`	Stable local ID
`platform`	Source platform
`source_type`	post, article, video, podcast, talk, repo, profile, comment, newsletter, press, milestone
`source_url`	Canonical URL when available
`raw_path`	Local raw evidence path
`captured_at`	Capture date/time
`capture_method`	One of the collection methods above
`auth_context`	public, owner_export, admin_export, manual, unknown
`terms_risk`	low, medium, high
`title`	Title or short label
`body_text_path`	Path to extracted text if present
`published_at`	Publication date
`creator_role`	author, speaker, host, guest, founder, maintainer, commenter, curator
`media_type`	text, image, video, audio, code, slide, mixed
`topic_tags`	Curated tags
`content_role`	acquisition, trust-building, proof, education, launch support, community, career signal
`metrics`	Platform-specific metrics object
`metric_confidence`	observed, owner-provided, estimated, unavailable
`evidence_confidence`	high, medium, low
`privacy_notes`	Redactions and sensitive field notes
`review_notes`	Human review notes and conflict flags

Metrics object fields are optional and must not be invented:

views
impressions
reads
listens
likes
reactions
comments
replies
reposts
shares
saves
clicks
subscribers
followers
engagement_rate

Creator Presence Dossier

The dossier should synthesize evidence into a durable Markdown artifact:

research/creator-presence/<slug>.md

Required sections:

Identity and current public promise.
Career timeline.
Platform map.
Core themes and expertise claims.
Proof assets.
Signature formats and recurring content patterns.
Audience and community signals.
Product/company connections.
Gaps, contradictions, and stale positioning.
Evidence register.
Next collection tasks.
Recommended next skills.

LinkedIn should be a first-class source in the dossier, but not the only source. The dossier should also ingest personal websites, GitHub, podcasts, talks, newsletters, and product docs when present.

LinkedIn Baseline

LinkedIn personal evidence should start from owner-provided exports and manual snapshots:

Profile data export.
Articles, shares/posts, comments, reactions, rich media, recommendations, skills, positions, education, and connections where the user chooses to provide them.
Public profile page snapshots only when accessible without login or when saved manually by the user.
Post screenshots or copied post URLs/text supplied by the user.

LinkedIn analytics should be treated as unavailable unless owner-provided. Company/page analytics and official APIs may be documented as a later authorized lane, not a baseline dependency.

Edge Cases

A platform export contains private contacts or sensitive messages. The skill must ask the user to redact or exclude those files before analysis.
A public page blocks automated capture. The skill must stop and request a manual snapshot rather than trying to bypass access controls.
A source has content but no metrics. The audit should proceed with a metric-gap warning.
A source has metrics but no content body. The audit should classify performance without claiming content-quality causes.
A creator has multiple names, handles, companies, or career eras. The dossier should preserve aliases and timeline boundaries.
A manual snapshot is stale. The evidence register must show capture date and confidence.
Evidence from different platforms conflicts. The dossier should flag the conflict rather than forcing one narrative.

Test Plan

Validate new skill files exist in mirrored Claude/Codex pack directories.
Run skill metadata scans for version fields and broken skill references.
Run targeted rg checks for collection methods, output paths, and LinkedIn baseline language.
Create a small fixture set under /tmp or documented sample paths representing manual LinkedIn export rows, public page captures, and RSS items; verify the schema can represent all records without platform-specific hacks.
Check git diff --check.

Acceptance Criteria

The creator-foundation pack documents a matrix-first expansion path beyond YouTube.
The shared evidence schema supports LinkedIn, websites/blogs, newsletters, podcasts, GitHub, and social platforms without requiring paid APIs.
The LinkedIn baseline uses owner export and manual/public snapshots, not logged-in scraping.
The creator presence dossier can be maintained entirely in Markdown in a git repo.
Future platform-specific skills can declare which matrix rows and schema fields they support.
Missing metrics and missing bodies are explicit evidence gaps.

Open Questions

Which LinkedIn export fields should be included by default versus redacted by default?
Should the first implementation include parser scripts for LinkedIn exports, or only skill contracts and manual templates?
Should the creator presence dossier be one artifact per creator or one directory per creator with separate evidence registers?
Should public web capture be implemented with Playwright, trafilatura, or both in the first execution phase?

Assumptions & Risks

[from spec] The work expands the creator-foundation pack (formerly creator-media) beyond YouTube. Risk: if the goal changes to only LinkedIn, the matrix may be broader than needed.
[from codebase] Existing skills are Markdown contracts mirrored for Claude and Codex. Risk: implementation must preserve mirror parity.
[from codebase] Existing YouTube workflows store raw evidence before synthesis. Risk: skipping raw evidence would weaken audit quality.
[confirmed] The first implementation target is a platform capability matrix plus shared evidence schema. Risk: downstream platform skills should wait until this foundation exists.
[confirmed] LinkedIn is the first priority platform. Risk: LinkedIn's restricted platform access makes owner export/manual capture essential.
[confirmed] Free, open-source, and manual evidence collection should be the default. Risk: some metrics and automation will remain unavailable.
[confirmed] Web scraping may be used only for safe public evidence capture, not as the foundation. Risk: overusing scraping could create brittle or noncompliant workflows.
[inferred] A repo-backed creator presence dossier should include public/professional evidence and optionally repo context when relevant. Risk: private planning context may leak into a public-facing dossier if scopes are not marked clearly.